DESCRIPTION: Radiotherapy is the principal treatment modality for advanced cervical cancer, but local control is frequently not achieved. The failure rate may be reduced by treating high-risk patients with more intense therapies including higher doses of radiation, chemotherapy, and/or surgery. However, there is no well-established predictor to identify patients whose high risk for failure justified the increased morbidity of more aggressive therapy. The investigators seek to identify those at high risk early, such that more aggressive treatment can be rendered that may improve outcome. Quantitative tumor volume and enhancement pattern analysis based on sequential MRI examination were shown to provide very early signals of failure in plot studies. Tumor size and dynamic enhancement pattern judged by the MRI prior to radiation therapy and temporal changes during the early course of radiation therapy appear to be sensitive predictors of tumor response; consistent with the notion that tumor blood supply and or oxygenation status strongly influence radiation response. The overall goal of this project is to test the hypothesis that MR-based measurements predict the likelihood of tumor control in patients with advanced cervical cancer treated by conventional radiation therapy. This will be achieved by three specific aims: (1) further develop, test, and refine predictive metrics of advanced cervical cancer radio-responsiveness based on contrast enhanced MRI and MR-based tumor volumetry, (2) apply MRI in a clinical population through their course of therapy and correlate tumor response with image-based metrics, and (3) determine predictive value (positive and negative) of MRI-based metrics. On completion, this project will provide a clinically validated MR protocol for prediction of tumor radio-responsiveness in advanced cervical carcinoma treated with radio-therapy. A prognostic index using MRI in a clinical setting to identify the high-risk patients who require more aggressive multi-modality therapy will be developed. The pixel signal distribution within the entire tumor between the radiosensitive and resistant groups will be further defined using multi-spectral and multi- temporal analysis, and characterized to discern subgroups contributing to treatment failure within the heterogeneous tumor.